University of Nebraska - Lincoln
DigitalCommons@University of Nebraska - Lincoln
Papers in Behavior and Biological Sciences
Papers in the Biological Sciences
1982
Response strategies in the radial arm maze:
Running around in circles
Sonja I. Yoerg
University of Massachusetts - Amherst
Alan C. Kamil
University of Massachusetts - Amherst
Follow this and additional works at: http://digitalcommons.unl.edu/bioscibehavior
Part of the Behavior and Ethology Commons
Yoerg, Sonja I. and Kamil, Alan C., "Response strategies in the radial arm maze: Running around in circles" (1982). Papers in Behavior
and Biological Sciences. 2.
http://digitalcommons.unl.edu/bioscibehavior/2
This Article is brought to you for free and open access by the Papers in the Biological Sciences at DigitalCommons@University of Nebraska - Lincoln.
It has been accepted for inclusion in Papers in Behavior and Biological Sciences by an authorized administrator of DigitalCommons@University of
Nebraska - Lincoln.
Animal Learning & Behavior
1982, 10 (4), 530-534
Response strategies in the radial arm maze:
Running around in circles
SONJA I. YOERG and ALAN C. KAMIL
University of Mussachusetts, Amherst, Massachusetts
The effects of the size of the central arena on the use of response strategies by rats on an
eight~arm elevated maze were examined. The size of the central arena had no effect on
accuracy, but the use of adjacent arms increased significantly with a larger central arena,
regardless of the size of arena to which rats were first exposed. These results are interpreted
in terms of foraging efficiency.
Many experiments employing the eight-arm radialstrategies. On a 17-arm radial maze, rats were conmaze have shown that rats readily learn to make theirfined to the central arena for 20 sec after each
choices so as to avoid arms already visited (see Olton, choice. While in the free choice condition the disHandelmann, & Walker, 1981, for a review). Vir- tribution of turn magnitudes was highly skewed in
tually errorless asymptotic performance often occursthe direction of the smaller magnitudes, the conwithin 20 trials (Olton & Samuelson, 1976). It hasfinement procedure was sufficient to disrupt any rebeen demonstrated that neither intramaze cues norsponse strategies. Accuracy was unimpaired by the
odor trails direct choice behavior on the maze, indi-manipulation.
cating that it is the characteristics from the wider These results contradict what intuition would inenvironment (i.e., the extramaze cues) that supportform us, namely, that the simplest method of obperformance (Olton & Collison, 1979; Olton, taining all the food available on the maze is to
Collison, & Werz, 1977; Olton & Samuelson, 1976). choose only adjacent arms. It has been pointed out
The reason rats perform so accurately in the radial that such a "choice strategy could be stored in longterm memory, and the arms could be responded to
maze may be that this method successfully mimics
as they appeared and would not have to be indicertain foraging situations in the wild. Contrary to
the general assumption of most psychologists, thatvidually identified or remembered" (Olton & Samuelson,
1976, p. 112). A question which might be posed is
reinforcing an animal will increase the probability
whether
there are conditions under which such strateof that animal’s returning to the place where reinforcement occurred, the behavior of rats on thegies might be expected to develop.
radial maze demonstrates just the opposite (Olton, Roberts and Dale (1981) tested rats on an eight1981). This finding is less of an anomaly if placedarm maze for five massed trials per day. The results
in an ecological context. A predator whose food indicated that, across days and across trials within
requires time to replete must have an accurate mem-days, the percentage of choices of adjacent arms
ory for the food locations already visited and notincreased to over 90°70 for all rats. Since accuracy
return to those emptied locations, if it is to forageimproved on Trials 4 and 5 (those most characterized by the adjacent arm strategy), Roberts and Dale
efficiently (Gill & Wolf, 1977; Kamil, 1978).
One rather curious finding in most radial mazesuggest that the adoption of the strategy served to
studies is that the rats do not appear to use strate- minimize proactive interference.
gies. That is, there is no evidence for consistent The physical features of the radial maze itself
response chains or algorithms (e.g., Olton & Samuelson,may also affect the pattern of responding in the
1976), although subjects do show a small tendency maze. Optimal foraging theory predicts that animals
to turn in one direction or another upon leavingwill use foraging strategies that tend to maximize
an arm and usually skip at least one arm beforerate of food intake over a given period of time
entering the next. Olton et al. 0977) conducted a (Charnov, 1976; MacArthur & Pianka, 1966; Pyke,
more rigorous test of the importance of responsePulliam, & Charnov, 1977). Field and laboratory
research based on this prediction has yielded positive results (e.g., Krebs, Ryan, & Charnov, 1974;
Werner
& Hall, 1974). One way of maximizing rate
Thanks are due Russ Church, Dave Eckermann, and Dave
Olton for comments and suggestions. This research was supported
of intake is minimizing the time (and/or energy)
in part by funds received from the University of Massachusetts
expended in traveling from one food source to
Psychology Department. Reprint requests should be addressed to
another. Menzel (1973) demonstrated this phenomthe senior author at the Department of Psychology, University
enon in chimpanzees, and discovered that the aniof C.alifornia, Berkeley, California 94720.
Copyright 1982 Psychonomic Society, Inc.
530
0090-4996/82/040530-05500.75/0
RESPONSE STRATEGIES IN THE RADIAL MAZE 531
mals retrieved hidden food not in the order in which week. The time to completion of the task and the sequence of
it had been hidden, but, rather, by using a "least- arms visited were recorded. Prior to each test session, three 45-mg
distance" or "nearest-neighbor" type of strategy. Noyes precision food pellets were placed in the food cup at the
of each arm. The rat was then placed in the center of
Such a strategy undoubtedly reduced the time and end
the platform and a timer was started. The session ended when
energy spent in traveling between cache sites, thus the rat had either obtained all the food or made 16 choices,
increasing rate of intake.
or when 10 min had elapsed. A choice was defined as having
In the case of the radial arm maze, the differ- occurred when the rat had placed all four feet upon an arm.
the next 10 days, the rats in Group S-L were placed
ence in energetic cost between traveling to an ad- onFor
the large arena maze, while those in Group L-S were placed
jacent arm or traveling to an opposite arm is prob- on the small arena maze. The criteria for the termination of
ably negligible, given the small size of the central a trial were the same as for Condition 1.
arena. Most of the distance traveled is incurred on
the arms. However, if the size of the central arena
RESULTS
relative to the length of the arm were to be increased, strategies involving the choice of adjacent
There were no group differences in the accuracy
arms might be expected to develop. The present of choice of unvisited arms during either phase of
experiment sought to determine the effect of in- the experiment.
creasing the size of the central arena of the radial
Both groups learned the task rapidly. By Day 8
maze on the pattern of selection of arms still con- of Condition 1, all rats had cleared the maze within
taining food.
10 min during three consecutive sessions. Group S-L
made a mean of 8.3 errors over these 8 days, while
METHOD
Group L-S made a mean of 7.8 errors [F(1,8) < 1].
The groups also did not differ in asymptotic acSubjects
curacy. For the last 10 days before transfer (Days 18The subjects were 10 male Holtzman rats, approximately 150
days old at the start of the experiment. Some of the rats had
27 of Condition 1), the mean number of different
received up to 12 1-mA shocks during equipment testing unarms chosen within the first eight choices was 7.9
related to the present study. The subjects were housed individfor Group S-L and 7.7 for Group L-S IF(l,8)=
ually and were maintained at 850~0 free-feeding weight during
1.98, p > .20]. The mean number of arms chosen
the course of the experiment. Daily feeding took place within
in order to clear the maze for these same 10 days
I h after testing.
was 8.1 for Group S-L and 8.4 for Group L-S
Apparatus
IF(l,8) = 1.56, p > .25].
The apparatus was an eight-arm elevated radial maze modiAfter transfer, both groups continued to perform
fied from that of Olton and Samuelson (1976). The arms were
accurately.
The rats in Group S-L chose an average
131 cm long and 8.5 cm wide. The proximal end of each arm
of 7.8 novel arms during the 10 days of this conwas nailed to a central supporting column which stood 53 cm
above the floor. One of two octagonally shaped arenas condition, while the rats in Group L-S chose an average
structed of 3-mm-thick Masonite rested on the supporting column.
of 7.7 novel arms [F(1,8) < 1]. There were also
The large arena was 88 cm in diameter, while the small arena
was 34 cm in diameter. A block of wood, 20 cm high, was no group differences in posttransfer accuracy in
terms of the number of choices (8.4 in each group)
attached to each arm. The placement of these barriers was adjusted so that the distance from the edge of both the small and required to complete the maze [F(1,8) < 1].
large arenas to the barrier was always 88 cm. A plastic furniTo demonstrate that choice accuracy was not disture coaster, 4.5 cm in diameter, was placed in front of each
rupted
by the transfer of Group S-L to the large
block and served as a food cup. The lip of the cup was 1 cm
maze
or
of Group L-S to the small maze, a 2 x 2
high and prevented visual detection of food by the rat at a disanalysis of variance was conducted with group and
tance of more than 10 cm. The maze had no walls.
The maze was located in a large classroom, illuminated by maze type as the independent variables. The differoverhead fluorescent lights. Desks, chairs, shelves, two doors,
ences between groups in the number of novel arms
and a blackboard were present, providing abundant visual cues.
chosen in the first eight choices before and after
An experimenter was seated in a corner of the testing room
during all sessions. A manual timer with a sweep second hand transfer were nonsignificant [F(1,8) = 1.11, p > .30].
was used to time trials.
There was no effect for maze type IF(I,8) < 1] and
no significant interaction between maze type and
Procedure
group [F(1,8) = 1.70, p > .20].
During the first 3 days, all rats were allowed to explore the
Data on the elapsed time per trial were recorded
apparatus with the small arena for 10 min each day. The subto the nearest minute. Over the last 10 days of
jects were then assigned randomly to either Group S-L (small
arena first) or Group L-S (large arena first). The rats in each
Condition 1 and the 10 days of Condition 2, 31.0070
group were then given an additional 10-min session of adaptaof all trials were completed in 1 min, 54.5070 were
tion with the arena on which they would be tested during Concompleted in 2 min, 10.0070 were completed in 3 min,
dition 1. No food was present on the maze during adaptation.
and 4.507o were completed in 4 min or more. InspecFor the next 27 days, Group S-L was tested with the small
arena in place and Group L-S was tested with the large arena
tion of the mean elapsed time per group per conin place. The rats were given one trial per day, 7 days per dition revealed no consistent difference in the time
532
YOERG AND KAMIL
required to clear the large and small mazes. Group S-L
required means of 1.9 min per trial in Condition 1
and 2.4 min per trial in Condition 2. Group L-S
completed the task in means of 1.7 and 1.6 min
per trial in Conditions 1 and 2, respectively. Generally, the rats spent little time in the central area,
making their choices quickly.
If the rats on the large maze were employing a
strategy involving adjacent arms, then we would expect to find significantly more turn magnitudes of
+ 1 (e.g., a choice of arm 5, followed by a choice
of ,either arm 4 or arm 6) for these rats, regardless
of whether they ran on the large maze before or
after transfer. This expected result was found. The
percentage of ± 1 turn magnitudes for Group S-L
was 35.3°?0 on the small maze and increased to 69.10?0
on the large maze. Group L-S had a mean of 56.60?0
+ 1 turn magnitudes on the large maze. The mean
percentage fell to 38.10?0 when these rats were transferred to the small maze. The percentage of + 1
turn magnitudes on the large maze was greater than
that on the small maze, irrespective of group [F(1,8)
= 18.02, p < .01] (see Figure 1). The mean percer~tage of ± 1 turn magnitudes across groups on the
large maze was 62.80?0, and was as high as 97.1070
for two of the rats in Group L-S. For the small
maze, the mean percentage of ± 1 turn magnitudes
across groups was 36.70?0. There was no main effect for group IF(l,8) < 1] and no interaction between group and maze type [F(I,8) = 1.52, p > .20].
A similar analysis of the relative frequency of ± 2
turn magnitudes indicated that the rats made signif70
6O
SMALL
ARENA
~ 50
I0
|
|
=
.
| = = =
±3
±4
±1
±2
TURN MAGNITUDE
±3
t4
Figure 1. The distribution o! turn magnitudes on the i~ge
and small mazes, collapsed over groups. The dotted line on the
right-hand panel is the mean distribution trom the control animals in Experiments I and 2 (Echermann et ni., 19~0).
icantly more turns of this magnitude when on the
small maze, irrespective of group [F(1,8) = 8.35,
p < .05], although the effect was not as large as
that for turn magnitudes of ± 1 (see Figure 1). The
mean percentage of 22 turn magnitudes across
groups on the small maze was 33.90?o, with the highest percentage of any individual rat being 60.30?0.
For the large maze, the mean percentage of +2
turn magnitudes across groups was 21.30?0. No main
effect for groups was found [F(I,8) < 1], and there
was no interaction between group and maze type
[F(I,8)= 3.30, p > .10]. The distribution of all turn
magnitudes for pre- and posttransfer trials on small
and large mazes is presented in Figure 1. Sample
data from two rats in each group are given in the
Appendix.
DISCUSSION
Increasing the size of the central arena from 34
to 88 cm resulted in a distribution of turn magnitudes highly skewed in the direction of adjacent
arms. This effect was evident regardless of the order
of presentation of the large and small arenas, demonstrating that previous experience on the small maze
neither interfered with nor was necessary for the
observed change in response patterning. Furthermore, after 27 days of experience on the large maze,
the percentage of + 1 turn naagnitudes of Group L-S
during only 10 days of testing on the small maze
was not different from that of Group S-L on the
small maze. The behavior of’ Group S-L also changed
rapidly when tested in the large maze; the percentage
of + 1 turn magnitudes increased to the level exhibited by Group L-S at the end of Condition 1.
These results strongly suggest that increasing the
cost of choosing nonadjacent arms by enlarging the
central arena is sufficient to produce nonrandom
choice sequences. However, the high degree of accuracy obtained on both mazes, with or without the
use of the adjacent arm response strategy, shows
that mnemonic ability and strategy use were independent dimensions of performance in this experiment.
It might be instructive to compare the data obtained from the small maze with those of other
experiments using similar apparatus and procedure.
In terms of accuracy, the performance of our subjects was as good as or better than that reported
for rats in similar situations (e.g., Olton & Samuelson,
1976). Furthermore, the distribution of turn magnitudes reported by Eckermann, Gordon, Edwards,
MacPhail, and Gage (1980) in their nondrug groups
of rats is highly comparable to the distribution obtained from small maze trials in the present study
(see Figure 1). It is therefore likely that the results
presented here are general.
RESPONSE STRATEGIES IN THE RADIAL MAZE 533
REFERENCE NOTES
Bond (Note 1) and Olton (Note 2) have suggested
that the increased use of adjacent arms may be the
Bond, A. B. Personal communication, 1982.
result of the larger turning arc between adjacent arms 2.1. Olton,
D. S. Personal communication, 1982.
in the large maze. For example, rats in the small
maze may not be able to easily negotiate the relatively sharp angle required to enter a neighboring
REFERENCES
arm. There may be some optimal turning radius
(possibly determined by rat size and a mechanical CHARNOV, E. L. Optimal foraging: The marginal value theorem.
advantage of maintaining constant velocity) that TheoreticalPopulation Biology, 1976, 2, 129-136.
D. A., GORDON, W. A., EDWARDS, J. D., MAcPHA~L,
combined with distance traveled to affect the dis- ECKERMANN,
C., & GAGE, M. I. Effects of scopolamine, pentobarbital,
tribution of turn magnitudes. These ideas are not R.
and amphetamine on radial arm maze performance in the rat.
inconsistent with our results, particularly the rela- Pharmacology, Biochemistry, and Behavior, 1980, 12, 595-602.
tively frequent use of +2 turn magnitudes in the GILL, F. R., & WOLF, L. L. Non-random foraging by sunbirds
in a patchy environment. Ecology, 1977, $8, 1284-1296.
small maze.
KAMIL,
A. C. Systematic foraging by a nectar-feeding bird, the
Some rather weak evidence against these notions
Amakihi (Loxops virens). Journal of Comparative and Physiocomes from the behavior of drugged rats on the logical Psychology, 1978, 92, 388-396.
radial maze (Eckermann et al., 1980). One might KEENS, J. R., RYAS, J. C., & CHARSOV, E. L. Hunting by exexpect difficulty in making a turn to be a function pectation or optimal foraging7 A study of patch use by chickadees. Animal Behaviour, 1974, 22, 953-964.
of the velocity of the animal. In comparison with MACARTHUE,
l:l,. H., & PIANKA, E. R. On the optimal use of a
saline control animals, amphetamine-injected rats patchy environment.
American Naturalist, 1966, 100, 603-609.
cleared the maze more quickly, although their accu- MENZEL, E. W. Chimpanzee spatial memory organization. Science,
racy and pattern of arm selection remained unaf- 1973, 182, 943-945.
fected. If these rats were making more choices per OLTON, D. S., & COLLISON, C. Intramaze cues and "odor trails"
fail to direct choice behavior on an elevated maze. Animal
minute because they were running faster, then this Learning
1979, 7, 221-223.
would appear to be inconsistent with the notion OLTON, D.&S.,Behavior,
COLLISON, C., & WERZ, M. A. Spatial memory and
that the choice of nonadjacent arms on the small radial arm maze performance in rats. Learning and Motivation,
maze is due to some optimal or convenient turning 1977, 8, 289-314.
angle. It seems certain that the choice behavior of OLTON, D. S., HANDELMANN, G. E., & WALKER, J. A. Spatial
and food searching strategies. In A. C. Kamil & T. D.
rats on the radial maze is affected by an entire memory
Sargent (Eds.), Foraging behavior: Ecological, ethological, and
set of variables, such as distance traveled and turn- psychological approaches. New York: Garland, 1981.
ing angle, which have received little attention.
OLTON, D. S., & SAMUELSON, R. J. Remembrance of places
This experiment tested a common sense hypoth- passed: Spatial memory in rats. Journal of Experimental
Psychology: Animal Behavior Processes, 1976, 2, 97-116.
esis. Imagine a radial arm maze with a central arena PYKE,
G. H., PULL~AM, H. R., & CHARNOV, E. L. Optimal
1 km in diameter. The gains in efficiency which an foraging:
A selective review of theory and tests. Quarterly Readjacent arm strategy would confer are obvious. view of Biology, 1977, $2, 137-154.
Nonetheless, the phenomenon reported here suggests ROnERTS, W. A., & DALE, R. H. I. Remembrance of places
several intriguing possibilities for future research, lasts: Proactive inhibition and patterns of choice in rat spatial
memory. Learning and Motivation, 1981, 12, 261-281.
including the role of the ratio of arm length to WEnsEa,
E. E., & HALL, D. J. Optimal foraging and the size
arena size and the influence of response strategies
selection of prey by the bluegill sunfish (Lepomis macrochirua).
on memory when confinement procedures are used. Ecology, 1974, $$, 1042-1052.
Appendix
Raw Data From Two Rats From Group S-L and Two Rats From Group L-S During
the Last 10 Days of Condition 1 and the First 10 Days of Condition 2
Group S-L
Small Maze
Group L-S
Large Maze
Large Maze
R2
3
6
5
3
8
8
4
2
7
8
85
264
1 7
84 21 753
3 164 827
64 1 2857
64 2753
1
516 2741
3
27853 16
85
743
16
1528643
6 14 278345
76 83 1 254
87
1 234 56
256 781
34
81 3456
72
36
78245
1
1457 82316
7543 1682
83
264 71
5
145782316
1 76543
28
87654321
87654321
87654321
87654321
87654321
87654321
87654321
87654321
13456782
12345678
Small Maze
R8
87654321
87654321
76543218
87654321
56781234
12345678
12345678
12345678
87654321
65432187
534
YOERG AND KAMIL
Group S-L
Small Maze
Group L-S
Large Maze
Large Maze
R4
58136724
61345782
71345682
68234571
82345671
713456782
71345682
57123468
58234671
81345672
2186543217
78234615
82456713
76543218
65432187
87654321
87654321
823456781
81234567
78123456
67812345
67812345
84567123
81234567
12345678
78123456
56781234
834567812
56781234
67812345
Small Maze
R10
68124735
5 7 8 1 2 34 5 6
.57812346
:] 2345671
823567134
1683715824
148712365
15684723
36812475
14678235
Note-R4 and RIO were chosen because their choice patterning was clearly dependent upon the size of maze used. R2 and R8 were
the rats whose behavior was the least responsive to the manipulation. Numbers 1-8 refer to the arms of t~he radial maze.
(Manuscript received February 8, 1982;
revision accepted for publication May 21, 1982.)